SWDEV-102733 - [OCL-LC-ROCm] Cmake build Write CMakeLists.txt to enable building with and without the DK environment
- Disable LLVM object compression. Allows closed source LLVM binaries to run with the open source runtime
- Define the DEBUG macro for Debug build types
- Use the prepared amdgcn lib instead of the static lib when generating the .bc includes for the runtime.
Affected files ...
... //depot/stg/opencl/drivers/opencl/CMakeLists.txt#14 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/CMakeLists.txt#9 edit
... //depot/stg/opencl/drivers/opencl/runtime/CMakeLists.txt#6 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/CMakeLists.txt#8 edit
1) hipSetDevice sets a flag so that next call to hipCtxGetCurrent returns primary context on current device
2) hipCtxGetCurrent returns primary context on current device if TLS context stack is empty
3) hipCtxPopCurrent falls back to primary context on current device as default
4) hipCtxPushCurrent, hipCtxSetCurrent and hipCtxCreate reset the flag set in hipSetDevice
SWDEV-102733 - [OCL-LC-ROCm] Cmake build Write CMakeLists.txt to enable building with and without the DK environment
- Use the internal libelf instead of the system's when building with cmake.
- Don't include codecvt if WITH_LIQUID_FLASH is not defined
- Don't memset the device::Kernel::workGroupInfo_, there's a std::string in it. The constructor already zeroes all the fields.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/CMakeLists.txt#8 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_lqdflash_amd.cpp#21 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rockernel.cpp#25 edit
SWDEV-102733 - [OCL-LC-ROCm] Cmake build Write CMakeLists.txt to enable building with and without the DK environment
- Use the internal libelf instead of the system's when building with cmake.
- Don't include codecvt if WITH_LIQUID_FLASH is not defined
- Don't memset the device::Kernel::workGroupInfo_, there's a std::string in it. The constructor already zeroes all the fields.
Affected files ...
... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/CMakeLists.txt#8 edit
... //depot/stg/opencl/drivers/opencl/api/opencl/amdocl/cl_lqdflash_amd.cpp#21 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/rocm/rockernel.cpp#25 edit
SWDEV-2 - Change OpenCL version number from 2469 to 2470.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2217 edit
SWDEV-79278 - [OCL][PAL] use HwPipeBottom to query the start time of an event.
The reason is that if there are many dispatches run in a loop with profiling enabled, then it is possible that the HwPipeStart of an event can occur before the HwPipeBottom of the previous dispatch and this will affect the timing reported by profiling.
ReviewBoardURL = http://ocltc.amd.com/reviews/r/13242/diff/1/
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/paltimestamp.cpp#3 edit
SWDEV-2 - Change OpenCL version number from 2468 to 2469.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2216 edit
SWDEV-2 - Change OpenCL version number from 2467 to 2468.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2215 edit
SWDEV-95919 - Expose all counters PAL supports.
At the request from CodeXL, and with the counter block index list provided by CodeXL.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palcounters.cpp#13 edit
SWDEV-2 - Change OpenCL version number from 2466 to 2467.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2214 edit
SWDEV-2 - Change OpenCL version number from 2465 to 2466.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2213 edit
- Refactoring introduced a bug when user does not specify any target via
--amdgpu-target, but has an invalid target specified in
HCC_AMDGPU_TARGET. In this case the selection logic was defaulting to
gf803.
- Removed defaulting to any specific target if rocm_agent_enumerator
fails. hipcc will report this and die if linking was required.
Change-Id: I76131867049fef92331807dd19a926406dcc1d02
SWDEV-2 - Change OpenCL version number from 2464 to 2465.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2212 edit
SWDEV-95919 - Avoid flushing when PerfCounter is enabled.
To make sure PerfStart/dispatch/PerfEnd are in the same cmdBuffer
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.cpp#51 edit
... //depot/stg/opencl/drivers/opencl/runtime/device/pal/palvirtual.hpp#27 edit
Existing logic has a bug. If user specifies targetA via commandline
options, while enumerator returns targetB, hipcc will create a fatbin
containing targets targetA and targetB. enumerator should only be used
when no target is specified by user (commandline or env var).
Change-Id: I6da857f86860c0e671b5988cd858644a08f723b9
SWDEV-2 - Change OpenCL version number from 2463 to 2464.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2211 edit
SWDEV-2 - Change OpenCL version number from 2462 to 2463.
Affected files ...
... //depot/stg/opencl/drivers/opencl/runtime/utils/versions.hpp#2210 edit
foreman
Rahul Garg
Siu Chi Chan
Jenkins
Ben Sander
Maneesh Gupta
Aditya Atluri
Evgeny Mankov
sunway513
Yifan Sun
Evgeny Mankov